Running head: UNDERSTANDING SCAFFOLDING. Expanding an Understanding of Scaffolding Theory. Using an Inquiry-Fostering Science Program.

Transcription

1 Understanding Scaffolding 1 Running head: UNDERSTANDING SCAFFOLDING Expanding an Understanding of Scaffolding Theory Using an Inquiry-Fostering Science Program Hee-Sun Lee Graduate School of Education University of California, Berkeley Nancy Butler Songer School of Education The University of Michigan Acknowledgement: This material is based in part upon research supported by the National Science Foundation under grant REC Any opinions, findings, and conclusion or recommendations expressed in this publication are those of the authors and do not necessarily reflect the views of the National Science Foundation. This paper is based on the dissertation study conducted by the first author of this paper. An earlier version of this paper was presented at the Annual Meetings of the American Educational Research Association, San Diego, CA, April The authors gratefully acknowledge Joseph Krajcik, Betsy Davis, and Philip Myers for their support of this work. The authors also thank Mike Derhammer, Bette Diem, and their students who showed great enthusiasm and provided their perspectives throughout this study.

2 Understanding Scaffolding 2 Abstract This quasi-experimental study investigated several parameters of scaffolding theory in an 8- week, technology-enhanced, biodiversity curriculum. Two treatments were created to test whether or not written curriculum materials should gradually diminish the amount of scaffolding as students gain experience with explanation tasks in real-world contexts. Forty-eight students in two 5th/6th combined classes were assigned to the two treatments. The consistent support treatment provided three kinds of epistemic explanation scaffolds, e.g. exemplars, questions, and sentence starters, throughout the eleven inquiry situations in the curriculum. In the fading support treatment, these scaffolds were gradually withdrawn over the three curricular phases. Data included pre and post tests, written explanations, and post interview transcripts of selected students. Results demonstrate that both treatment groups exhibited pre to post gains in the knowledge about biodiversity and the ability to match given evidence to a claim. However, as the fading of explanation scaffolds occurred to the fading support treatment, the consistent support group outperformed the fading support group in formulating explanations from authentic data. High ability students appeared to benefit from consistent content support, while low ability students underutilized the explanation scaffolds provided in this study.

3 Understanding Scaffolding 3 Introduction One of the most challenging goals of the current science education reform is preparing students for lifelong learning (Linn & Muilenburg, 1996; NRC, 1996). In lifelong learning situations that occur outside classroom settings, students need to formulate, conduct, and sustain their own inquiry in real-world contexts (Minstrell & van Zee, 2000; NRC, 2000). Science class provides an ample opportunity for students to acquire and polish independent skills needed for real-world inquiry (Bybee, 2000). To support inquiry reasoning that is both meaningful to students and faithful to the scientific enterprise, the National Science Education Standards advocate the use of authentic science activities that are similar to those they [students] will encounter in the world outside the classroom, as well as to situations that approximate how scientists do their work (NRC, 1996, p. 78). Authentic inquiry is complicated for students to pursue because they often lack much of the domain-expertise required to effectively reason within real-world contexts (Lee & Songer, 2003). Despite epistemological and motivational advantages of using authentic inquiry (Edelson, 1998), the challenge of, for example, determining salient from irrelevant evidence makes it extremely difficult for students to carry out authentic inquiry as is common within the science community (Chinn & Malhotra, 2002; Kuhn, 1989). Research demonstrates that in any given reasoning situation, students may lack content knowledge, inquiry experience, technological resources, professional commitment, or community support (Bransford, Brown, & Cocking, 2000; Edelson, 1998). As a result, students seldom create meaningful inferences from data without adequate support (Keys, 1999). Scaffolding can be one effective instructional means to address these expert-novice differences (Quintana, Reiser, Davis, Krajcik, Fretz, Duncan, Kyza, Edelson, & Soloway, 2004).

4 Understanding Scaffolding 4 The scaffolding metaphor applies to instructional situations where students, following the guidance of the more knowledgeable other, become competent with academic tasks that are initially beyond their ability (Palincsar, 1998; Wood, Bruner, & Ross, 1976). The more knowledgeable other can successfully diagnose the complex needs of students at various stages of the intended learning and employ proper instructional strategies adaptively to their progress (Tabak, 2004). Recent theoretical advances in sociocultural and distributed cognitions allow the expansion of the more knowledgeable other: from human tutors (Wood et al., 1976) to teachers in classrooms (Palincsar & Brown, 1984), peers (Hogan, Nastasi, & Pressley, 2000), technological artifacts (Quintana et al., 2004), and curriculum materials (Cazden, 2001; Davis, 2003; Linn & Hsi, 2000). Stone (1998) points out that detailed mechanisms of fading in the scaffolding framework are not clearly understood. Bruckman (2000) suggests that answers to when, how, and for whom scaffolding can successfully fade may not be generalizable beyond the initial learning context. Pea (2004) calls for empirical evidence that can test the claims of scaffolding theory in which scaffolding theory involves specific formulations of what distinctive forms and processes of focusing, channeling, and modeling are integral to the development of expertise (p. 443). Pea (2004) also suggests such mechanisms should be examined for students with different capabilities on the intended learning. The study investigated fading mechanisms in scaffolding students development of scientific explanations from data they collected as part of an eight-week curricular program, BioKIDS: Kids Inquiry of Diverse Species (Songer, et al., 2002). The BioKIDS program, is designed to support middle school students active pursuit of inquiry problems in biodiversity. Two treatments were designed to test whether or not written curriculum materials should gradually diminish the amount of scaffolding as students gain

5 Understanding Scaffolding 5 experience with explanation tasks in real-world contexts. In the consistent support treatment, three types of explanation scaffolds, e.g. exemplars, questions, and sentence starters, were provided throughout the eleven inquiry situations in the BioKIDS curriculum. In the fading support treatment, these scaffolds were gradually withdrawn over the three curricular phases. In comparing these two treatments, this study addressed the two following questions: Question 1: How did the learning outcomes of the fading support group compare to those of the consistent support group? Question 2: How did students with different levels of expertise relative to content knowledge and explanation generation respond to fading as compared to consistent support? In the following sections, this paper discusses background literature related to (1) explanation-focused, real-world inquiry, (2) interventions to support students explanation building process, and (3) scaffolding. Explanation-Focused, Real-World Inquiry Situated cognition provides a theoretical framework for the analysis of what constitutes authentic activities in science class. Brown, Collins, and Duguid (1989) state that authentic activities represent ordinary practices of the culture (p. 34) where meanings and purposes are socially constructed through negotiations among present and past members (p. 34). A focus on selected ordinary practices of the scientific community leads to two possible emphases for authentic science activities for students. McDonald (2004) defines these two emphases as authentic learning and authentic science. In authentic learning, the science activities are situated in everyday life to prioritize students interests (Keys, 2001; Krajcik et a., 1998; Linn & Muilenburg, 1996; Linn & Songer, 1991). In authentic science, a priority is put on students experiencing a streamlined version of scientists research or data-gathering practices (Chinn & Hmelo-Silver, 2002; Edelson, 1998). These two emphases do not always contradict each other

6 Understanding Scaffolding 6 (NRC, 1996). However, it is often difficult to accomplish both goals with any one set of activities. Though the epistemological and motivational appeal of authentic science learning is understandable (Metz, 2000), using real-world problem contexts in science class presents many challenges. First, real-world science is inherently complicated. Unlike controlled experiments in the laboratory, real-world phenomena involve many variables that can influence the outcomes of investigations. Students need to assess the relative importance of each variable and prioritize major variables to make causal relationships with the observed outcomes. Prioritizing major variables is difficult and requires domain specific expertise (Lee & Songer, 2003). Second, students lack advanced knowledge and thinking skills necessary to address complicated real-world problems (Edelson, 1998). Differences between experts and novices are well documented including concept understanding (Anazi, 1991; Lewis & Linn, 1994), scientific reasoning (Clement, 1991; Kuhn, 1989; Schunn & Anderson, 1999), problem solving (Chi, Feltovich, & Glaser, 1981), explaining scientific text (Chi, Lewis, Reimann, & Glaser, 1989), and understanding scientific diagrams (Lowe, 1993). Most students have difficulty learning science within real-world contexts without explicit guidance in, for example, recognizing salient data or diagram components (Edelson, 2001; Metz, 2000). One program that studied the challenges in making 24 hr weather forecasting situations revealed the importance of creating authentic activities that (1) map to real-world contexts that are interesting to students, (2) emphasize problem situations that are more similar to instructional reasoning situations, and (3) provide guidance in determining salient from irrelevant evidence (Lee & Songer, 2003). Bridging cognitive differences between scientists and students is also found in White s (1989) intermediate abstraction. Taken together, translated real-world inquiry contexts retain the major aspects of scientific inquiry scientists perform, but provide

7 Understanding Scaffolding 7 simplified language, easier access to information, and support in determining salient information in order to allow novices to be more successful at complex inquiry (Lee & Songer, 2003). In scientific inquiry, scientists create, revise, or reject scientific knowledge including theories, models, laws, and explanations (Schwab, 1962). The goal of teaching science through authentic inquiry is not only to polish students inquiry skills but also to enhance their ability to acquire, apply, and refine their knowledge using scientifically valid methods and techniques. Student artifacts such as models and explanations are good indicators of knowledge development (Spitulnik, Stratford, Krajcik, & Soloway, 1998; Sandoval, 2004). Since a major goal of scientific inquiry is to gain explanatory insight into the physical world (Hempel, 1966; Kuhn, 1970; NRC, 1996), a curricular goal to focus on students development of explanations presents similar thinking processes inherent in inquiry reasoning as performed by scientists (Sandoval, 2004). In addition, learning how to explain has other educational benefits. Coleman (1998) suggests that explaining enables one to reason more logically and scientifically, promotes understanding scientific theories within domains, fosters understanding about why problems are formulated as they are, and, most important, clarifies what needs to be explained (p ). However, school science generally does not provide ample explanation opportunities for students (Chi, Leeuw, Chiu, & Lavancher, 1994; Kuhn, 1993). As a result, most students have difficulty formulating logically consistent explanations that connect to their scientific knowledge (Bell & Linn, 2000; Bransford, et al., 2000; Butcher & Kintsch, 2001; Kuhn, 1989) especially with authentic data (Keys, 1999). Interventions to Support Students Explanation Building Process In order to facilitate scientific writing, instruction needs to help students work fluently in both rhetorical and content spaces (Bereiter & Scardamalia, 1987). In the rhetorical space,

8 Understanding Scaffolding 8 students need to think how to develop a plausible, convincing argument. In the case of developing scientifically valid arguments, rhetorical support would help students recognize the importance of (1) having a logical consistency between claims and explanations (Bell & Linn, 2000), (2) providing relevant data to justify their claims (Keys, 1999), and (3) communicating ideas clearly to the reader (Scardamalia, Bereiter, Brett, Burtis, Calhoun, & Lea, 1992). Examples of instructional supports designed to work in the rhetorical space may look like, Make sure you include data to justify your claim, and List three pieces of evidence to support your main idea. Such rhetorical prompts that do not connect to specific content can be called content-free or content-lean. In the content space, students need to make domain-specific connections between evidence and their claims (Kuhn, 1989). To assist this aspect of scientific writing, relevant scientific content can be prompted using questions such as What does hemoglobin transport? (Chi, et al., 1994) or using sentence starters such as The normal predator-prey relation is the factor in the relationship that has changed is (Sandoval, 2003). Direct content prompting can be particularly useful in real-world problem solving where students have difficulty focusing on salient features (Lee & Songer, 2003). Directly prompting content can also be an effective strategy for novice students with weak domain expertise (Sandoval, 2004). In comparing the effects of content prompts and rhetorical prompts on college students writing, Butcher and Kintsch (2001) found out that [u]se of content prompts results in clear and immediate benefits to time spent in the writing process stages and to the quality of the text that is produced for novice science writers, prompting consideration of and decisions about content results in the most powerful and impressive benefits to the writing process and written text quality (p. 317). Benefits of content prompting relate to students increased ability to generate

9 Understanding Scaffolding 9 ideas (Klein, 2000; Patterson, 2001). The importance of content support in scientific writing is not surprising as Millar and Driver (1987) point out that: [i]t appears that what children notice, what they will do and the interpretations they give depend on the conceptions they use. These in turn depend on their prior knowledge in a particular context or domain of experience. Thus, a pedagogy which focuses primarily on the learning of processes may be fundamentally misguided (p. 51). This implies that inquiry skills such as explanation and argumentation cannot be properly developed without supporting students acquisition of domain-specific knowledge (Hodson, 1988; Kuhn, 1992). Even though molecular biologists know what constitutes a good scientific explanation, it is possible that they may not write scientifically valid explanations in the context of high energy physics. Similarly, students cannot write scientifically valid and sophisticated explanations if they do not possess knowledge that can provide a theoretical foundation for their explanations. Recalling relevant content knowledge is not enough for students to formulate scientifically valid explanations. In fact, students should also be familiar with epistemic forms of scientific inquiry situated in a specific disciplinary context (Collins & Ferguson, 1993). Sandoval (2004) further elaborates the relationship between general epistemological commitments and discipline specific paradigms. The criteria that scientists have for what counts as a good theory in their discipline depend upon the questions that they find important to answer, but also conform to more general epistemic criteria, such as coherent causal mechanisms, parsimony, and so on. General `epistemological commitments entail beliefs about what counts as valued and warranted scientific knowledge, and lead to the development of investigative strategies that can produce such knowledge. The canonical strategy of controlling variables across experiments is valued because it allows for the isolation of causal relationships, an epistemic goal. (p.347-8)

10 Understanding Scaffolding 10 Therefore, explanation building can be more effectively supported if scaffolding focuses on both relevant science content and explanation templates to demonstrate how to achieve causal coherence and establish evidentiary support. Simplistic templates, such as proving key terms, may not be enough. For example, Cavallo, McNeely, and Edmund (2003) asked ninth grade students to write an essay by saying that In your summary, include an explanation of how CHEMICAL REACTIONS may be related with the following terms: atoms compounds chemical change (p. 589). Cavallo et al. (2003) found that merely giving students key terms without adequate instruction encouraged students to retain misconceptions and misuse the terms. Even though research concludes that students formulation of explanations needs adequate support, it is not clear whether students benefit more from content-specific or from rhetorical support in real-world inquiry contexts. In some studies, explanation-fostering supports focus on either content space or rhetorical space, and their impacts are compared to no support conditions (Chi, et al., 1994; King, 1994; McNeill, et al., 2004). Other studies suggest that to maximize the impact of explanation supports, researchers need to design interventions that address both spaces (Sandoval & Reiser, 1997). Whether students need content-free or specific prompts may be dependent upon a range of factors including student knowledge, explanation experience, and scientific complexity of the situation that needs to be explained. As for the specificity of scaffolds, Davis (2003) found that middle school students can reflect more productively from generic reflection prompts than from specific reflection prompts that dictate what to reflect. Scaffolding In order to differentiate scaffolding from other types of instructional strategies, Stone (1998) identifies four key characteristics. A learning context establishes an academic task that is initially unachievable by students. The teacher provides adequate support commensurate to

11 Understanding Scaffolding 11 students progress (adaptive). To provide adequate support to learners with varying abilities, the teacher has a repertoire of support strategies and methods (diverse). The teacher s support should gradually fade to assure the transfer of responsibility to students (fading). Two distinctive aspects of the scaffolding concept are its emphasis on fading (Guzdial, 1994; Pea, 2004; Stone, 1998) and its sensitivity to students developing knowledge and ability to perform the task (Palincsar & Brown, 1984). Fading refers to the gradual reduction of support by the more knowledgeable agent in successful tutor-tutee (Wood et al., 1976), mother-child (Wertsch & Stone, 1985), teacher-student (Fleer, 1992; Flick, 2000), or expert-apprentice relationships (Brown, et al., 1989). In the design-based research paradigm where curriculum materials or technological artifacts primarily guide student learning, the definition of fading is less distinct. For example, Pea (2004) argues that curricular or technological scaffolds without apparent fading mechanisms may not be considered as scaffolding but as distributed intelligence. Scaffolding-minded interventions can address fading either as a part of the student s learning process or as an explicit, active intervention strategy. One fading approach in distributed learning environments is to leave fading entirely to the users of scaffolds such as teachers and students. In many of the current science education research studies focusing on students development of complex thinking, scaffolds embedded in written curriculum or technological innovations do not fade during the intervention period (for example, Davis, 2003; Quintana et al., 2004; Sandoval, 2003; White & Frederiksen, 1998). Tabak (2004) suggests a synergistic approach to addresses scaffolding at multiple levels, from design to implementation of scaffolds in order to take the ecological complexity of real classrooms into account. Another fading approach is for curricular materials or technological artifacts to fade scaffolds systemically. In this case, fading can be based on the diagnosis of student progress (Anderson, Boyle, Corbett, &

12 Understanding Scaffolding 12 Lewis, 1990) or on predetermined pedagogical decisions that assume the development of expertise (McNeill et al., 2004). However, clear guidelines for fading mechanisms are not present in the literature and thus need greater specification (Stone, 1998). One explanation for this lack of specificity is the understanding that every child s learning is unique and depends on many instructional factors including leaner characteristics (McNamara, Kintsch, & Songer, 1996), task characteristics (Lee & Songer, 2003), and interactivity with the learner (Bruckman, 2000). Learner Characteristics As the teacher s decisions are sensitive to the progress of each individual student, effective fading mechanisms for guiding students in complex reasoning tasks should differ between high and low achieving students (Pea, 2004). This sensitivity may be even more pronounced for elementary and early middle school learners (Resnick & Klopfer, 1989; Kuhn, Black, Jeselman, & Kaplan, 2000). Metz (1997) suggests the importance of higher-order reasoning tasks for younger students, [m]ost problematic, the targeting of purportedly elementary science processes for the first years of schools with a postponement of the integrated practice of goal-focused investigations until the higher grades results in decomposition and decontextualization in the teaching and learning of scientific inquiry. As a consequence, young children engage in science activities such as observation and categorization apart from a rich goal structure or overriding purpose, a practice which is detrimental from cognitive, motivational, and epistemological perspectives. (p. 152) Metz (1995) further argues that higher-order thinking skills can be acquired by students if instruction is designed to address the weaknesses of the students. To support students independent inquiry, Metz (2000) designed a curriculum that scaffolds domain-specific knowledge, knowledge about empirical inquiry, domain-specific methodologies, data analysis,

13 Understanding Scaffolding 13 and tools. Results indicate that even elementary students as young as second grade are able to engage in part of the scientific inquiry process such as designing controlled experiments. A similar case can be made for low achieving students. White and Frederiksen (1998) implemented reflective assessment into an inquiry cycle model in the ThinkerTools curriculum to foster qualitative understanding of the Newtonian Mechanics concepts. Even though students with all abilities benefited from the ThinkerTools curriculum, White and Frederiksen (1998) noted the benefits for low achieving students. In another case using several instructional modules in the Science, Technology, and Environment in Modern Society (STEMS) project, Zohar and Dori (2003) found that higher-order thinking skills can be taught to students with all abilities and the gap between high and low achieving students can be narrowed with judicious scaffolding. One of the interesting differences between high- and low-achieving students is in what level of support optimizes their performance. In reading comprehension, McNamara, Kintsch, and Songer (1996) found that high achieving students work best with a version of the scientific text that requires inference-building by the learner, while low achieving students need a version of the text with all inferences provided. A similar result is found in an explanation writing task that uses real-world problems. Rivard and Straw (2000) discovered that high ability students do better when they work independently while low to average ability students benefit from peer discussion. These studies suggest that though students with all abilities can benefit from the same treatment (White & Frederiksen, 1998; Zohar & Dori, 2003), tailoring support to meet the differential needs of students optimizes student achievement (Davis, 2003). Task Characteristics To understand scaffolding, Bruckman (2000) notices that the content of the help you receive matters, but the context in which that support is situated is also of great importance (p. 330). The learning context is defined by the characteristics of the academic task given to the

14 Understanding Scaffolding 14 student (Doyle, 1983) as well as the type of interactions the student has with the social and physical resources (Bruckman, 2000; Blumenfeld, Mergendoller, & Swarthout, 1987). Task characteristics affect both the content of scaffolds and the duration of the scaffolds. If a learning situation allows students to employ the same strategies during a series of academic tasks, it is likely that students internalize these strategies from repeated uses, allowing the fading of support (Palincsar & Brown, 1984; McNeill et al., 2004). However, real-world learning contexts often do not provide consistent contexts in which students can recognize and internalize strategies (Chinn & Malhotra, 2002). For example, Lee and Songer (2003) studied students justifications for their 24-hr forecasting in several real-world weather situations. The real-world forecasting task asked students to make predictions on maximum and minimum temperatures, clouds, precipitation, and wind direction. The quality of students justifications for 24-hr forecasts depended upon how closely the real-world situations mapped onto their content understandings about weather systems rather than the accumulation of forecasting experiences. Therefore, scaffolds that directly reduce task complexity through channeling and focusing are necessary, at least until students possess domain expertise to address the task effectively (Pea, 2004). Interactivity with the Learner Distributed cognition increasingly recognizes that scaffolding can occur through various means. First, scaffolding can occur through verbal interaction with the teacher as in the case where teachers observe student s progress and prescribe needed support (Wood et al., 1976). In Palincsar and Brown s (1984) reciprocal teaching, teachers modeled a set of specific comprehension strategies such as summarization with students then very gradually turned over these comprehension strategies to students as competence was realized. One of the challenges in replicating this approach was the necessity of a great deal of teacher attention towards each student, and accurate diagnosis of student progress.

15 Understanding Scaffolding 15 Second, peers can scaffold each other (Bruckman, 2000). Scardamalia et al. (1992) showed the benefits of a student-generated communal database that allows students to share and critique each other s ideas and projects. The student-generated database improved the quality of acquired knowledge, written products, and questions generated for further scrutiny (Scardamalia et al., 1992). The effectiveness of peer scaffolding is explained by the shared communication and learning experience between students (Rivard & Straw, 1999) that is not always present between teacher and student (Bruckman, 2000; Hogan, et al., 2000). However, Blumenfeld, Marx, Soloway, and Krajcik (1996) warn that productive collaborative group work is not a guarantee. The effects of group work depend on how the group is organized, what the tasks are, who participates, and how the group is held accountable (Blumenfeld et al., 1996, p. 37). Third, scaffolding can be delivered through technological resources. In some cases, technologies can include multiple layers of support from which students can choose (Guzdial, 1994; Linn, Clark, & Slotta, 2003). In other cases, technologies guide learners based on the record of their responses to prompts on screen (Anderson, et al., 1990). Sometimes technologies provide fixed scaffolds (Davis & Linn, 2000; Sandoval, 2003). Scaffolding embedded in technological tools can access a large number of students with standardized support. However, even with an exemplary scaffolding design, many scaffolding tools are underutilized by students who lack metacognitive awareness of their own learning progress. Fourth, scaffolding can be delivered through written curriculum materials (Cazden, 2001). Some curricula are deliberately designed to improve the learner s ability to conduct independent inquiry. For example, White and Frederiksen (1998) use reflective assessment at the end of each inquiry cycle to help students reflect on their inquiry process in the modeling of Newtonian Mechanics concepts. Compared to mentoring, the advantage of scaffolding through written curriculum materials is that many learners can be reached simultaneously to achieve

16 Understanding Scaffolding 16 specific learning goals set by curriculum designers. The disadvantage is that written curriculum materials are relatively insensitive to individual variations towards the expected competence. The Research Context This section describes the learning context in which two scaffolding conditions were manifested as well as the students and teachers who participated in this study. The BioKIDS Program The BioKIDS program (Songer et al., 2002) is a technology-enhanced, inquiry-focused biodiversity curriculum for fifth and sixth grade students. Students are guided through eight weeks of activities that utilize their own data about the biodiversity of their schoolyard towards higher order thinking in science. The topics of biodiversity, food webs, and ecology are the central scientific concepts. The curriculum sequence reflects an inquiry cycle of engage, explore, analyze, and synthesize. In the engage phase, students are introduced to the biodiversity concept, data collection methods, and technological resources including CyberTracker (Parr, Jones, & Songer, 2002) and Critter Catalog (Espinosa et al., 2002). CyberTracker is software that enables students to record and organize animal sightings systematically. Critter Catalog is a web resource that provides rich information on local animals including appearance, habitat, food, predator-prey relationships, reproduction, human interaction, and endangerment. After the engage phase, students use their own data to explore and analyze a deeper conceptual understanding of scientific concepts such as biodiversity, including an examination of animal abundance and species richness. Students also evaluate how microhabitats in their schoolyard support the animals. In the synthesize phase, each student choose an animal and gather information from Critter Catalog to evaluate the ecological need of their focus animal. Students then determine whether their animals can survive in the schoolyard by comparing what their animals need and what their schoolyard can provide. Students create food webs using the

17 Understanding Scaffolding 17 animals they investigated in the previous part to explore concepts such as consumers, producers, decomposers, herbivores, carnivores, and omnivores. In addition, students learn about energy flow, interdependence, and interrelationships among the organisms in their food webs. Participants This research was conducted in a K-8 school in the Midwestern United States. Due to the open philosophy in this school, students were familiar with hands-on, interest-based, studentdirected projects. Forty-eight students in two 5th/6th mixed classes participated in this study. Mr. Moss taught one class of 28 students, and Ms. Boyle taught the other class of 21 students. Two thirds of the students were Caucasian. The rest consisted of seven multi-racial, five African American, two Hispanic, one Asian, and one American Indian student. Results of a district-wide standardized test indicated that these students performed slightly higher than the state average on the state science test. Mr. Moss has been teaching in this school for ten years after he acquired a teaching certificate in elementary education. Ms. Boyle has been teaching in this school for more than twenty years and had a teaching certificate in general and elementary education. Both teachers had prior experiences with technology and innovative curricula. Both teachers understood the purpose of this study and expressed their support. All BioKIDS classes were observed by the first author of this paper. Often, the researcher acted as a participant observer helping students conduct investigations and clarifying directions on the worksheets. Two Treatments This study was designed to provide empirical evidence on fading mechanisms in a distributed learning environment where written curricular worksheets primarily guided individual students investigations. Unlike most research on scaffolding that compares student

18 Understanding Scaffolding 18 outcomes between with and without scaffolds, this study compared student outcomes and artifacts between with consistent scaffolds and with fading scaffolds. The academic task of focus was for students to formulate explanations to justify their claims using evidence collected from their inquiry investigations. Three types of explanation scaffolds were identified from the literature. First, questions (Q), similar to those used in Chi et al. (1994) and Butcher and Kintsch (2001), oriented students to focus on a small number of salient features for problem solving. For example, questions provided in classifying animals based on observable characteristics include, Do they have external skeletons? and How many legs do they have? To compare explanations across a range of biodiversity problems, four salient features were emphasized through questions (Q) in each problem. Second, exemplars (E) gave students an idea of how the salient features mentioned in questions (Q) could be incorporated into an explanation. An exemplar for the classification problem is, I think a beetle and an ant can be grouped together because they have external skeletons and six legs. These data show that both of them are insects. Third, Sentence starters (S) asked students to fill in their claim and justification as shown in this example, I think and can be grouped together because [list relevant data or information] Sentence starters for claims were different across inquiry problems, but were consistent relative to evidence generation, e.g. because [list relevant data or information] Questions, exemplars, and sentence starters used in this study are listed in Tables 1, 2, and 3, respectively Insert Tables 1, 2, 3, & 4 Here Using these three types of explanation scaffolds, two treatment conditions were established in terms of how fading of the scaffolds occurred over time. As shown in Table 4, both treatments offered exemplars, questions, and sentence starters in Phase I (Problem 1 to 3).

19 Understanding Scaffolding 19 In the fading support treatment, exemplars were eliminated in Phase II (Problem 4 to 7), and then questions were eliminated in Phase III (Problem 8 to 10). In the consistent support treatment, all three types of scaffolds were provided throughout all phases. This fading mechanism was determined to respond to three assumptions: (1) students need to see and recognize what is expected at the beginning of the task (exemplars); (2) channeling and focusing can reduce the complexity of the real-world task (Pea, 2004; Sandoval, 2003); and (3) scaffolds that address more cognitively challenging aspects of the task need to remain longer (Lee & Songer, 2003). There are two possible projections on the outcomes of this study. The first possibility is that students gaining experiences with the explanation building tasks allow content-specific scaffolds to fade. Students with consistent content support on the other hand may consider scaffolds as sources for information throughout the tasks, and therefore do not develop important skills for writing explanations such as prioritizing salient evidence. The second possibility is that a series of complicated real-world inquiry situations presents different challenges to students, which may result in denying students full engagement with the explanation tasks. Comparable (or higher) achievement of the fading support group to the consistent support group can imply that the intentional fading of scaffolds is beneficial for students development of independent explanations. If the consistent support group prevails, this research can provide contextual insights towards how and why fading of content scaffolds becomes ineffective. Data Collection & Analysis Before the eight-week BioKIDS curricular program began, all students took a two part pretest (content test and claim-evidence test). Students were teamed in groups of four to start the first BioKIDS activity. Each student group was assigned to one of the two treatments based on prior biodiversity knowledge measured on the content pretest and prior explanation ability measured on the claim-evidence pretest. Each treatment had six student groups. There was no

20 Understanding Scaffolding 20 statistically significant difference between the two treatment groups on the content test, t(45) = 1.35, p =.25, or the claim-evidence test, t(43) =.05, p =.70. During the BioKIDS curricular program, students engaged in eleven inquiry problems where the two treatments were embedded. In the first BioKIDS activity, students observed animals in their schoolyard. Following this activity, students developed individual explanations without scaffolds to address the pre-treatment problem, Among animals I saw in the schoolyard, which animals can be grouped together? Right after the first explanation, students were given a second opportunity to explain the same problem on a new worksheet that featured all three types of scaffolds (Problem 1). From Problem 2 to 10, students received the scaffolds as shown in Table 4. After the treatments ended, all students wrote explanations about the post-treatment problem that did not provide any explanation scaffolds. All students then took the content posttest as well as the claim-evidence posttests. In addition, nine students from each treatment were interviewed. Interviewees represented various levels of knowledge and explanation ability. The sections below describe data sources and analysis procedures. Content Test The content test addressed knowledge about the biodiversity concepts in the eleven inquiry problems. This test had 19 multiple-choice and 3 short-answer items taken from released standardized tests such as the National Assessment of Educational Progress (NAEP), the Third International Mathematics and Science Study (TIMSS), and the Michigan Educational Assessment Program (MEAP). All responses were coded as either correct (1 point) or incorrect (0 point). The maximum score for the content test was 22. To demonstrate the pre to post improvement, effect sizes were used. To detect treatment effects, repeated measure MANOVA s were performed using time (to account for pretest to posttest increase) as a within-subjects variable and treatment type as a between-subjects variable.

21 Understanding Scaffolding 21 Claim-Evidence Test The claim-evidence test consisted of five items that measured students ability to use given evidence to justify their claims. Each item asked students to make a claim about a biodiversity problem and explain their claim. Item stems were similar to sentence starters (S) used in the treatment. For instance, Figure 1 shows a question about the adaptation of Viceroy Butterflies. This question presents pictorial as well as textual evidence for students to determine whether Viceroy Butterflies would be eaten by predators. In scoring the claim, one point was given for a scientifically correct claim. In scoring the explanation part, one point was given for each of two correct pieces of evidence. The maximum score for the claim-evidence test was 32. The intercoder reliability was.95. To show pre to post gains and treatment effects, the same statistical techniques were used as described in the previous section. Explanations Insert Figure 1 Here All of the eleven inquiry situations were based on students investigations in the BioKIDS curriculum. The investigations were completed either individually or in groups, but students were asked to explain their claims individually. Students explanations were expected to meet the following criteria: Explanations need to include evidence relevant to the problem. Evidence in the explanation should be scientifically valid to justify the claim. A claim can be better justified if a larger number of scientifically valid warrants are presented. Following these criteria, students explanations were coded in terms of the number of warrants, the number of valid warrants, and the validity ratio. A warrant was implied in each pair of a claim and a piece of evidence, allowing multiple warrants to exist in an explanation. After a total

22 Understanding Scaffolding 22 number of warrants was counted in an explanation, each warrant was examined to determine scientific validity. A warrant was as coded as invalid if: (Invalid data) Data are not properly measured or cited. (Irrelevant data) Data are irrelevant to the problem. (Inconsistent data) The connection between the claim and data is inconsistent. The number of valid warrants was calculated by counting how many warrants were scientifically valid in each explanation. The maximum number of valid warrants in each explanation was 4. The validity ratio, ranging from 0 to 1, was calculated by dividing the number of valid warrants with the number of warrants. This validity ratio measure emphasized the quality of the explanation as a whole, while the number of valid warrants emphasized the quantity of valid warrants in the explanation. This explanation coding process was described using Mary s explanation: I think a roly-poly and a spider can be grouped together because they each have eight legs and they re both invertebrates. They both eat insects. These data show that they are arachnids. In her explanation, she implicitly made four warrants: A roly-poly and a spider can be grouped together because they both have eight legs. A roly-poly and a spider can be grouped together because they both are invertebrates. A roly-poly and a spider can be grouped together because they both eat insects. A roly-poly and a spider can be grouped together because they both are arachnids. The first warrant is invalid because a roly-poly has more than eight legs. The second warrant is valid. The third warrant is irrelevant to the problem because the inquiry problem asks students to focus on physical characteristics not behaviors. The fourth warrant is invalid because a roly-poly does not belong to arachnids. The number of valid warrants in this explanation is 1, and the validity ratio is.25.

23 Understanding Scaffolding 23 Students explanations were coded separately by two independent coders. Intercoder reliability in all of the inquiry situations ranged from.89 to.99. To compare explanations of the fading and the consistent treatment groups during the course of this study, repeated measures MANOVA s were conducted on the number of valid warrants and the validity ratio, using time as a within-subjects factor. Time representd a series of pre treatment (without scaffolds), Phase I, Phase II, Phase III, and post treatment (without scaffolds). To create a collective score in each phase, explanation scores over the problems in the same phase were averaged. Interviews Interview data provided students perspectives on how they used explanation scaffolds while they were formulating explanations. During an interview, each student was presented with his/her explanations about the following three problems: Problem 2 from Phase I: Are and the same species? Problem 7 from Phase II: What kinds of adaptations are used by my animal to survive in its habitat? Problem 8 from Phase III: Can my animal live in my schoolyard? All of the eighteen interviewed students were asked to rank the easiest and the most difficult problems to explain and tell why. In addition, students who received fading support were shown missing scaffolds and asked whether and how they liked to change their previous explanations. Each interview took about 20 minutes. All of the student interviews were recorded on audiotapes and transcribed. Interview segments are used in this paper to support the authors interpretation of research findings. Research Question 1: Comparing Two Treatments This section compares the fading support treatment and the consistent support treatment on the content test, the claim-evidence test, and the explanations.

24 Understanding Scaffolding 24 Content Test & Claim-Evidence Test The content test assessed whether students acquired knowledge about biodiversity. The claim-evidence test assessed student ability to make scientifically valid links between the given evidence and their claims. As shown in Figure 2, both treatments demonstrated significant pre to post gains on the content test, F(1, 41) = 14.32, p <.001, as well as on the claim-evidence test, F(1, 39) = 32.81, p <.001. In addition, the consistent support group demonstrated larger gains than the fading support group on both the content test, ES =.48 vs. ES =.28, and the claimevidence test, ES =.64 vs. ES =.44. There was no interaction effect between treatment and time (pre to post). These results indicate that both groups, to a similar extent, acquired (1) the biodiversity knowledge that was necessary to formulate explanations and (2) the ability to match given evidence to a claim. Explanations Insert Figure 2 Here Figure 3 (a) illustrates the trajectories of the two treatment groups on the number of scientifically valid warrants in explanations. In the pre-treatment problem, the fading support group, without scaffolds, included a larger number of scientifically-valid warrants than the consistent support group. Both treatment groups improved from the pre-treatment explanation to the Phase I explanations. This finding indicates that all three types of content scaffolds were beneficial to students. In Phase II, the removal of exemplars (E) in the fading support treatment was associated with a slight decline in the number of valid warrants. However, the number of valid warrants increased in the consistent support treatment group in the same phase. The decline of the fading support group continued as both questions (Q) and exemplars (E) were withdrawn in Phase III, while the consistent support group continued to improve. In the post-treatment

25 Understanding Scaffolding 25 explanation, both treatment groups experienced slight declines from their Phase III explanations as students developed explanations without any types of scaffolds. These trajectory changes over time were statistically significant, F(4, 144) = 7.29, p <.001. In addition, a statistically significant interaction effect existed between time and treatment, F(4, 144) = 7.75, p <.001, which reflects the consistent support group s improvement over time versus the fading support group s decline. Figure 3 (b) shows the trajectories of the two treatment groups on the validity ratio. The validity ratio of both treatment groups improved over time, F(4, 144) = 12.51, p <.001. There was neither treatment difference, F(1, 36) =.34, p =.56, nor interaction effect, F(4, 144) = 1.36, p =.26. As students experiences with explanation building tasks increased, both treatment groups appeared to evaluate each piece of evidence more carefully to determine whether it could support their claim. However, it is noticeable that while the consistent support group continued to improve the validity ratio during the entire treatment period, the fading support group did not improve after Phase II. This finding may suggest that questions (Q) were essential for students to make scientifically valid connections. Without questions (Q), students might have worked harder to select what data to emphasize and how to make connections between the data and their claims. Discussion Insert Figure 3 Here Students, regardless of treatment type, demonstrated pre to post gains in knowledge about biodiversity as well as the ability to match the given evidence to their claims. However, the consistent support group came to make a larger number of scientifically valid warrants than the fading support group. The improvement of the consistent support group and the decline of the

26 Understanding Scaffolding 26 fading support group coincided with the gradual withdrawal of scaffolds in the fading support treatment. Student interview data provide insights to understand this outcome. Responding to diverse needs of students. Two important skills students need to formulate explanations in real-world inquiry contexts are (1) focusing on salient evidence of the problem (Lee & Songer, 2003; Sandoval, 2003, 2004) and (2) building consistent, logical arguments between data and a claim (Bell & Linn, 2000; Keys, 2001; Sandoval, 2003). Each student s progress towards perfecting these abilities is potentially unique. Some students may need content support that identifies salient features all the time, while others can formulate explanations on their own after only a few explanation building practices. In this study, consistent content support better responded to the diverse needs of students who could be in different stages of acquiring explanation skills than the fading content support. Students with consistent support appeared to spend more time thinking about connections between particular data and their claims, instead of examining data from various sources and determining which to use in their explanations. In addition, the predetermined fading order might not adequately reflect students progress towards independent explanation building as well as their preference for the type of the content support they needed. The short treatment duration might also contribute to the unsuccessful fading of content support. Even though scaffolds were presented on the worksheets, whether and how individuals used them depended on students own assessment of what they needed. This study suggests that some students ignored scaffolds on their worksheets if they were confident about their answers for the problem. Ted in the consistent support group said that he did not consider exemplars (E) or questions (Q) because he knew the answer immediately after reading the problem: I: Did you read the example? Ted: No, I don t think I read the example